ZHT

ZHT: A Light-weight Reliable Persistent Dynamic Scalable Zero-hop Distributed Hash Table

• PDF
• Authors
  • Tonglin Li
  • Xiaobing Zhou
  • Kevin Brandstatter
  • Dongfang Zhao
  • Ke Wang
  • Anupam Rajendran
  • Zhao Zhang
  • Ioan Raicu
• Year: 2013
• Tags
  • DHT
  • HPC
  • Distributed systems

Notes

A supercomputer is a device for turning compute bound problems into I/O bound problems

— Ken Batcher

• Designed for high performance computing and supercomputers
• Use case has roughly 1:1 ratio of servers and clients
• Beyond a DHT: also data availability and uptime
• Roughly a key-value API
  • insert, remove, lookup, and additionally append
• How
  • Partition the key space into a ring
  • Replicate slices of the key space
  • Nodes enter and are given a key space adjacent to an existing one on the ring to replicate their partition
    • This key space can later be split if needed
    • On unexpected node exit, the replicated node provides availability, and asks others to replicate it
  • Many partitions/nodes per physical system with a single “manager”
• Built a distributed file system (FusionFS)
• Outperformed memcached in their tests, esp as the number of nodes grew